Monitoring indoor environmental conditions in office buildings using a sustainable Agentic RAG-LLM system

• Indoor Environmental Conditions impact health, productivity, and energy use. • Thermal comfort monitoring requires diverse data and intelligent analysis. • BIM integration with real-time data is often underutilized in thermal systems. • ThermalComfortBot uses GenAI to improve comfort and energy optimization. • Case study shows ThermalComfortBot achieves 94% Accuracy and high efficiency. Indoor Environmental Conditions (IEC) play a crucial role in determining the health, productivity, and overall building performance of employees, as well as their energy consumption. Key parameters, such as temperature and humidity, are not only vital for thermal comfort but also offer opportunities to enhance energy efficiency when effectively monitored and managed. Accurate Thermal Comfort Monitoring (TCM) remains challenging to achieve because it requires the integration of diverse data sources and intelligent analysis, particularly in light of evolving global energy and sustainability standards. Although Building Information Modeling (BIM) is increasingly being adopted to manage complex building data, its integration with real-time sensor inputs remains vastly underutilized. Existing thermal monitoring systems are often development-intensive, require significant domain expertise, lack Natural Language (NL) interaction capabilities, and are not inherently adaptable, necessitating frequent technical upgrades. These limitations give rise to pressing concerns about long-term scalability, usability, and sustainability. To address these limitations, this study introduces ThermalComfortBot , an integrated Information System (IS) powered by Generative Artificial Intelligence (GenAI). ThermalComfortBot utilizes open-source technologies, including Large Language Models (LLMs) and Agentic Retrieval-Augmented Generation (RAG), to enhance thermal comfort and support energy optimization in buildings. The system integrates Building Information Modeling (BIM), sensor data, and external datasets to generate actionable insights, delivered through both textual explanations and graphical visualizations. This system utilizes flexible and adjustable LLMs that are guided by principles of sustainability, thereby making them cost-efficient, scalable, and practical for a diverse range of organizational environments. In a real-world case study, ThermalComfortBot outperforms traditional RAG-LLM, achieving 94% accuracy, 92% precision, and 89% recall, enhancing comfort and efficiency.